In this paper, we use the recently proposed wet paper codes and introduce a new approach to passive-warden steganography called Perturbed Quantization. In Perturbed Quantization, the sender hides data while processing the cover object with an informationreducing operation that involves quantization, such as lossy compression, downsampling, or A/D conversion. The unquantized values of the processed cover object are considered as side information to confine the embedding changes to those unquantized elements whose values are close to the middle of quantization intervals. This choice of the selection channel calls for wet paper codes as they enable communication with nonshared selection channel. Heuristic is presented that indicates that the proposed method provides better steganographic security than current JPEG steganographic methods. This claim is further supported by blind steganalysis of a specific case of Perturbed Quantization for recompressed JPEG images.
KeywordsPerturbed quantization, wet paper code, adaptive, security, steganography
MOTIVATIONThe primary goal of steganography is to build a statistically undetectable communication channel (the famous Prisoner Problem [1]). In order to embed a secret message, the sender slightly modifies the cover object to obtain the embedded stego object. In steganography under the passive warden scenario [2,3], the goal is to communicate as many bits as possible without introducing any detectable artifacts into the cover object. Attempts to give a formal definition of the concept of steganographic security can be found in [4][5][6]. In practice, a steganographic scheme is considered secure if no existing attack can be modified to build a detector that would be able to distinguish between cover and stego images with a success better than random guessing [7].One possible measure to improve the security of steganographic schemes for digital media is to embed the message in adaptively selected components of the cover object [8][9][10], such as noisy areas or segments with a complex texture. However, if the adaptive selection rule is public or only "weakly dependent on a key", the attacker can apply the same rule and start building an attack. It is then a valid question whether the adaptive selection improves steganographic security at all. An interesting example of a scheme where adaptive pixel selection in fact decreased its security is the recent surprising result of Westfeld [11].This problem with adaptive steganography could be remedied if the selection rule was determined from some side information available only to the sender but in principle unavailable to the attacker. For example, imagine the situation when the sender has a raw, uncompressed image and wants to embed data into its JPEG compressed form. Can the sender use his side informationthe uncompressed image -to construct a better JPEG steganography? We can attempt to select for embedding those DCT coefficients whose unquantized values lie "close to the middle" of quantization intervals. Intuitively, p...